Clinical value of neuroimaging indicators of intracranial hypertension in patients with cerebral venous thrombosis

Between January 2012 and December 2017, we prospectively registered 70 patients presenting with acute CVT to our institution. Inclusion criteria were age > 18 years and CVT diagnosed by neuroimaging (CCT and CT-Angiography or MRI and MR-Angiography). Of these, 26 patients who completed the MRI study protocol were included in the cohort (supplementary Fig. 5). 26 age- and sex-matched healthy volunteers without neurological disease underwent the study MRI to serve as controls.

General exclusion criteria were contraindications against MRI (ferro-magnetic implant, pregnancy, claustrophobia), critical illness (e.g. mechanical ventilation, severely altered consciousness), preexisting conditions causing intracranial hypertension (idiopathic intracranial hypertension, hydrocephalus) and denial to participate. We further excluded CVT patients showing space occupying hemorrhage, edema or venous infarction with midline-shift or lateral ventricle compression (see below) [3] and patients with symptoms suggestive of IH, in whom CSF pressure measurement, ophthalmological examination and/or ocular MRI were unavailable.

After diagnosing CVT, patients underwent cerebral and ocular MRI and, upon clinical suspicion of IH, CSF pressure measurement and additional ophthalmologic examination.

All 26 patients were part of a previous cross-sectional assessment of IH predictors in acute CVT [3].

We classified CVT patients as having intracranial hypertension based on the following predefined criteria [3]:

As described before [3], we chose papilledema for its 95–100% specificity for IH [13, 14]. Given reports of unilateral papilledema in IH, we rated both uni- and bilateral optic nerve head protrusion as an indicator for IH [21]. Patients were defined as IH negative when they lacked criteria of the compound classifier.

CVT patients underwent baseline cerebral MRI at inclusion (3 Tesla MRI, Magnetom TRIO or 1.5 Tesla MRI, Magnetom Avanto, Siemens Healthcare, Erlangen, Germany) including 2D T1, 2D T2, 2D time-of-flight (TOF) venography, 3D fluid attenuated inversion recovery based on sampling perfection with application-optimized contrasts using different flip angle evolution (FLAIR SPACE), 2D susceptibility-weighted imaging (SWI), and 2D high-resolution T2-weighted half-Fourier acquisition single-shot turbo spin-echo (HASTE) sequences as previously described [19, 22, 23]. Supplementary Table 1 gives sequence details. We complemented dedicated orbital imaging and contrast-enhanced (ce-) vascular imaging at 3 T soon after CVT diagnosis, as described before [3]. Controls underwent the same MR protocol at 3 T as patients, yet without application of Gd contrast agent. CVT follow-up comprised the same MR protocol and neurological examination.

If not contraindicated, we acquired ce-T1W magnetization-prepared rapid acquisition with gradient echo (MPRAGE; supplementary Table 1) venography after acquision of the other sequences, approximately 4 min. after intravenous injection of 0.1 mmol/kg Gadoteridol (ProHance, Bracco Imaging, Milan, Italy), available in 24 patients. In two patients, we analyzed ce-CT-venography scans acquired after intravenous injection of 70 ml Iomeprol (Imeron 400, Bracco Imaging, Milan, Italy) at a flowrate of 3 ml/s, followed by 60 ml saline. CT-venography was started manually with inflow of contrast agent into the internal jugular vein using visual bolus tracking, as described before [3] using Siemens Somatom Definition AS and Definition Flash CT scanners (Siemens Healthcare, Erlangen, Germany); tube voltage 100 kV, exposure 140 mAs, collimation of 64 × 0.6 mm, helical mode, reconstruction of 0.75 mm isotropic datasets.

MRIs were rated by four experts with longstanding experience in neuroimaging analysis (blinded readers: TD and NL > 10 years, SM > 20 years in neuroradiology; FS > 10 years in cerebral MRI research) who interpreted anonymized data using standard imaging software (IMPAX EE R20 DeepUnity Diagnost, Dedalus HealthCare GmbH, Bonn, Germany). TD and SM analyzed venographies blinded to clinical characteristics, complementary morphological data and rating performed by NL and FS and vice versa.

Two Neuroophthalmologists with > 10 years (MR) and > 25 years (WL) experience in ophthalmology performed funduscopy, being unblinded to the CVT diagnosis but unaware of MRI and LP related data.

Physicians performing lumbar puncture were blinded to ophthalmologic and neuroimaging findings.

We identified CVT-associated cerebral edema in 3D FLAIR SPACE and hemorrhage using susceptibility-weighted imaging as described before [22]. In addition, on axial and coronal slices we searched for IH causes not attributable to thrombotic material, assessing any midline shift, lateral ventricle compression or dislocation of the temporal horns [3].

TD systematically determined the location and number of thrombosed intracranial veins and sinus on all available time-of-flight (TOF) and ce-MPRAGE venographies [24] as described before [3]. One point was assigned to each thrombosed part of the following 18 predefined segments (1–18 points): rostral, mid and dorsal segments of the superior sagittal sinus (SSS), transverse and sigmoid sinus, basal veins of Rosenthal, great vein of Galen, straight sinus and cortical veins draining into the SSS including the superior and inferior collaterals of Trolard and Labbé.

TD and SM independently rated thrombus recanalization on follow-up ce-MRV as described before [3](supplementary Table 2), using the grading proposed by Aguiar de Sousa [25] and Qureshi [26]. Readers resolved disagreement by consensus. We analyzed the influence of recanalization extent (any degree, partial (i.e. Aguiar de Sousa grade ≥ 2A, see supplementary Table 3) and complete recanalization (Qureshi grade III, Aguiar de Sousa grade 3)).

NL rated the presence of optic disc protrusion [3], bulbar flattening and optic nerve tortuousity [17] depicting the bulbi and optic nerves (Fig. 1; supplementary Fig. 1), pituitary tissue was graded using 3-dimensional FLAIR SPACE [22] according to Yuh et al. (grade 1, normal appearance through grade 5, empty sella; grade 3 representing abnormal partially empty sella; supplementary Fig. 2)[16]. Lateral and fourth ventricle size (supplementary Figs. 3 and 4) was determined using standardized measurements on axial T1w images as previously described [14].

FS analyzed ONSD (Fig. 1) and frequency of enlargement > 5.8 mm [18, 19] using HASTE-sequence as described before [19, 23]. In short, we placed a circular region-of-interest marker 3 mm behind the eyeball, perpendicular to longitudinal extension of the optic nerve. ONSD was defined as the border of highest contrast of signal intensity change between CSF and the surrounding optic nerve sheath.

Continuous data are given in mean ± standard deviation (SD). We assessed cross-sectional and longitudinal differences of continuous variables by simple and dependent t-test and categorical data by chi² or Fishers’ test. We determined sensitivity and specificity, positive and negative predictive values (PPV, NPV) and odds ratios (OR) of neuroimaging signs. Ordinal variables were analyzed by Mann Whitney U (cross-sectional) or Wilcoxon rank sum tests (longitudinal). We determined associations of continuous, categorical and ordinal variables by Kendall’s tau.

We included 26 CVT patients (21 female, 77.8%) and 26 age- and sex-matched controls. Patients’ demographics are presented in Table 1. Average time from CVT diagnosis to inclusion was 7.3 ± 8.7 days. 23 patients underwent follow-up ce-MRI after 184.9 ± 44.7 days, in the majority 5–7 months after CVT diagnosis. Later imaging was performed in three patients after 245 to 335 days. Of these, one patient (P7) was IH positive and two (P4 and P12) IH negative.

12/26 patients (46.2%) had IH (supplementary Fig. 6). We classified 14 patients by CSF pressure measurement (11 patients had elevated (42.9 ± 10 cmH₂O) and three normal CSF pressure (16.3 ± 1.5 cmH₂O)). Among 9 patients undergoing neuroophthalmologic examination, we classified three patients by the lack of papilledema (6 were already classified by CSF pressure). We classified 9 out of 23 patients by ocular imaging (one showing, 8 lacking papilledema; the remaining 14 patients were already classified by CSF pressure and ophthalmologic examination).

In this cohort, no patient had unilateral papilledema. Controls did not show papilledema at repeated MRIs 26.2 ± 10.9 days apart.

Until follow-up, IH symptoms resolved in most patients under treatment and greatly improved in P15 and P22 (Table 1). Patient P13 developed delayed IH, despite recanalization under anticoagulation (Aguiar de Sousa Grade 2B).

Several often intertwined mechanisms increase the IH risk in CVT: First, the amount and localization of thrombus predispose to IH [3, 8, 9, 27]. Second, depending on collateralization, even local thrombosis may increase endovenous pressure. Third, the pressure dependend CSF resorption via glymphatic pathways is reduced by lower trans-vessel pressure e.g. when thrombosis involves Pacchioni granulations [28]. Fourth, although rare, dural arterio-venous fistula causing or complicating CVT may increase intravenous and intracranial pressure [29‐31]. Fifth, local space occupation due to edema, infarction and hemorrhage following venous stasis may cause rise of intracranial pressure [32].

To classify IH, we relied on elevated CSF pressure [12] and if unavailable, on papilledema on ophthalmological examination and/or optic disc protrusion on MRI. The number of positive neuroimaging findings correlates with CSF pressure [33]. In addition, the prevalence of papilledema correlates with the number of IH neuroimaging indicators, increasing from 2.8% in patients presenting at least one sign to 40% in patients with ≥ 4 IH indicators [34].

With nearly one in two patients (46%) developing IH during acute CVT, IH prevalence was comparable to previous reports [1, 2]. IH rarely developed delayed [4], affecting two patients with progressive thrombosis 6 and 38 days after CVT diagnosis.

Our prospective CVT cohort appears closer to real-life in-hospital conditions than previous retrospective case–control cohorts selected by proven IH [14, 17], limiting their mutual comparability. Our results support the diagnostic value of ONSD enlargement regarding IH in acute CVT [18, 19]. In line with Dong et al. [14], pituitary grade differentiated CVT patients with IH and controls. The other neuroimaging indicators were less of avail. Although more common in patients with IH, optic nerve tortuousity had low sensitivity in identifying IH in acute CVT. Ocular bulb flattening frequency, pituitary grade, and—contrasting Dong et al. [14]—lateral and IV^th ventricle size did not discriminate CVT patients with versus without IH at baseline.

Morphological changes may appear at different CVT stages and show a diverging dynamic. Given the 1–2 week latency from CVT diagnosis to inclusion in our cohort, we cannot exclude misclassification of patients who developed IH and associated morphological effects later. Despite elevated CSF pressure, five patients in our cohort did not show papilledema at inclusion. Papilledema in CVT mostly develops within one to two months [7]. The limited positive predictive value of neuroimaging findings might thus stem from the early CSF pressure measurement. In addition, the lack of papilledema in patients not undergoing CSF pressure measurement might have underestimated IH frequency.

Little data is available on the usefulness of neuroimaging findings as monitoring parameters during the later CVT course, e.g. to guide the duration of anticoagulation in case of incomplete recanalization. Reversibility of neuroimaging indicators appeared heterogeneous, more frequently occurring in ocular findings in our cohort. The frequency of ONSD enlargement, bulbar flattening and pituitary grade decreased significantly, and reversibility was associated with IH classification. Yet, despite significant regredience, ONSD enlargement did not regress to baseline values in 8/8 IH patients compared to controls and patients without IH. Pituitary grade even increased in patients with IH.

As IH symptoms resolved in most patients, we estimate ongoing IH at follow-up is less probable. Patient P13 developed delayed IH despite advanced recanalization. Showing markedly improving IH symptoms but lacking follow-up LP, we cannot exclude ongoing IH causing slight residual blurred vision in P15 and intermittent tinnitus in P22.

Duration of IH, tissue elasticity and the time needed for normalization might affect the remission of IH neuroimaging signs. In idiopathic intracranial hypertension (IIH) patients showing papilledema, MRI findings suggestive of IH persisted after resolution of both symptoms and papilledema [35]. In CVT, papilledema was reversible on average after 6 months [7]. In vivo intrathecal infusion tests [36] and post mortem experiments [37] demonstrate the interrelation between ONSD enlargment and short-term CSF pressure elevation. ONSD reversiblility was incomplete after high pressure exposure [37]. In our cohort, we ascribe the persistent ONSD enlargment to prolonged IH during acute CVT.

At follow-up, partially empty sella in patients with IH remained significantly more frequent compared with controls, however not reaching significance compared with CVT patients without IH. Although IH treatment might reverse empty sella [38], pituitary compression may outlast due to structural changes of the sellar region, as in IIH [35].

Regredience of bulbar flattening frequency did not differ in patients with versus without IH, potentially due to the limited cohort size. Longitudinal changes of inner CSF spaces and IH at baseline were not significantly associated in our cohort. In line, the size of inner CSF spaces neither predicted IH nor differed before versus one month after treatment of CVT [39].

In the range of previous reports [40], depending on the grading used, 33% (Aguiar de Sousa) to 70% (Qureshi) showed complete and > 70% at least partial recanalization of all initially affected vessels. Several neuroimaging findings were non-significantly associated with partial recanalization of all previously thrombosed vessel segments (Aguiar de Sousa grade ≥ 2A). There was no correlation with presence or absence of complete recanalization alone.

Incomplete recanalization might predispose to longstanding IH, occurring in 10% after CVT [4]. In contrast, recanalization is associated with both normalization of hemodynamics, cerebral metabolism and positive outcome [25, 40‐43]. Given variable collateralization networks, both the extent of recanalization necessary to prevent IH and the degree of venous flow restoration necessary for regredience of imaging alterations in CVT remain unclear.

Strengths of our study include the prospective CVT cohort, reflecting real-life in-hospital conditions, which we analyzed by dedicated, systematic rating of thrombosis and recanalization. We internally validated test performance measures of neuroimaging signs by comparing IH subgroups versus their separate matched controls. Our results offer an orientation regarding the evolution of neuroimaging findings after CVT.

Our study is limited by its size and monocentric design, and also by exclusion of patients lacking both CSF pressure measurement and ophthalmologic examination or ocular MRI. Second, our cohort possibly overrepresents patients with visual symptoms and, excluding critically ill patients, our results might not apply to severe CVT. Third, for ethical reasons, we renounced follow-up CSF pressure measurement to confirm IH reversibility. We think this is negligible as IH symptoms resoved in all patients. Only one patient (P9) showed asymptomatic optic disc protrusion on follow-up MRI. Our results need replication, ideally in a larger prospective cohort.

The suspicion of CVT indicates emergent brain imaging work-up. Complementing ce-MR-venography, IH neuroimaging sings offer high specificity classifying IH. Unfortunately, the temporal course of IH and neuroimaging findings in CVT are highly variable and symptoms may precede morphological effects of IH. This carries the risk of misclassification and explains the limited positive predictive value of IH neuroimaging indicators. Thus, neuroimaging surrogates may support estimating the IH risk in CVT but appear insufficient to replace CSF pressure measurement.

Pausing anticoagulation for spinal tap carries the risk of progressive thrombosis. This requires evaluating the clinical presentation in synopsis with IH risk factors, e.g. thrombus location, extension and collateralization [3, 9, 11, 27]. Consequently, individual IH pretest probability should guide the decision to perform LP. Longitudinal ONSD assessment may be a helpful alternative, given the ease of sonographic bedside assessment.

Further, IH neuroimaging findings persisting after CVT or appearing delayed in asymptomatic patients may cause uncertainty regarding the need and duration of anticoagulation, clinical follow-up frequency and invasive CSF pressure measurement. In asymptomatic CVT patients, persisting IH indicators should promote non-invasive diagnostic steps before-hand, such as ophthalmological examination (e.g. optical coherence tomography, fundoscopy, ONSD sonography) [7, 13, 15]. In patients with a sustained suspicion of IH, we suggest determining CSF pressure to guide further therapy, e.g. continued anticoagulation or acetazolamide [44].

The optimal duration of anticoagulation is currently elusive [44] as 3/4 of patients recanalized within 3 months [40], but recanalization may occur as late as 800 days after CVT [42]. Therefore, patients showing incomplete recanalization need individualized counseling regarding anticoagulation duration, weighing the chances of further recanalization versus bleeding risk.